Automatically select appropriate model for a set of models

Share:

Description

Runs through a model selection algorithm to determine the best model in a given set

Usage

1
2
3
rank.models(data, ..., nested = F, bootstrap = F, model.type = "ssm",
  alpha = 0.05, robust = F, eff = 0.6, B = 50, G = 1e+05, freq = 1,
  seed = 1337)

Arguments

data

A vector, data.frame, matrix, or gts object with 1 column.

...

Different ts.models to be compared.

nested

A bool that indicates whether the ts.model objects are nested within a large object given within the list. If not, the a full model will be created.

bootstrap

A bool that is either true or false to indicate whether we use bootstrap or asymptotic By default, we use asymptotic.

model.type

A string indicating whether the model should be a "ssm" or "imu".

alpha

A double that indicates the level of confidence for the WV CI.

robust

A boolean that indicates whether to use robust estimation.

eff

A double between 0 and 1 that indicates the efficiency for the robust estimation.

B

A integer that contains the amount of bootstrap replications

G

A integer that indicates the amount of guesses for caliberating the startup.

freq

A double that represents the frequency between observations.

seed

A integer that is used to set a seed for reproducibility.

Details

The models MUST be nested within each other. If the models are not nested, the algorithm creates the "common denominator" model.

To supply the models, enter them as: AR1()+WN(), AR1(), 3*AR1()

Any parameter that you wish to use must then be specified. e.g. to specify nested, you must use nested = T. Otherwise, it the function will stop.

Due to the structure of rank.models, you cannot mix and match AR1() and GM() objects. So you must enter either AR1() or GM() objects.

Value

A rank.models object.

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.